[COFF] Re: Butler Lampson's 1973 Xerox PARC memo "Why Alto?"

[COFF] Re: Butler Lampson's 1973 Xerox PARC memo "Why Alto?"

[Noel Chiappa]

    > From: steve jenkin

    > What struck me reading this is the estimated price (~$10K) to build an
    > .. [ a note elsewhere says $4,000 on 128KB of RAM. 4k-bit or 16-kbit
    > chips? unsure ]

16K (4116) - at least, in the Alto II I have images of. Maxc used 1103's
(1K), but they were a few years before the Alto.


    > I believe the first "PDP-11" bought by 127 at Bell Labs was ~$65k fully
    > configured

I got out my August 1971 -11/20 price sheet, and that sounds about right. The
machine had "24K bytes of core memory .. and a disk with 1K blocks (512K
bytes ... a single .5 MB disk .. every few hours' work by the typists meant
pushing out more information onto DECtape, because of the very small disk."
("The Evolution of the Unix Time-sharing System"):

 11,450	Basic machine CPU + 8KB memory
  6,000	16KB memory (maybe 7,000, if MM11-F)
  4,000	TC11 DECtape controller
  4,700	TU56 DECtape transport
  5,000	RF11 controller
  9,000	RS11 drive
  3,900	PC11 paper tape
-------
 44,050

(Although Bell probably got a discount?)

The machine later had an RK03:

  https://minnie.tuhs.org/cgi-bin/utree.pl?file=V1/u0.s

but that wasn't there initially (they are 2.4MB, larger than the stated
disk); it cost 5,900 (RK11 controller) + 9,000 (RK03 drive).

Also, no signs of the KE11-A in the V1 code (1,900 when it eventually
appeared). The machine had extra serial lines (on DC11's), but they weren't
much; 750 per line.

    > Why the price difference?

Memory was part of it. The -11/20 used core; $9,000 for the memory alone.

Also, the machine was a generation older, the first DEC machine built out of
IC's - all SSI. (It wasn't micro-coded; rather, a state machine. Cheap PROM
and SRAM didn't exist yet.)

	Noel

[COFF] Re: Butler Lampson's 1973 Xerox PARC memo "Why Alto?"

[Larry Stewart]

[-- Attachment #1: Type: text/html, Size: 11151 bytes --]

[COFF] Re: Butler Lampson's 1973 Xerox PARC memo "Why Alto?"

[Clem Cole]

[-- Attachment #1: Type: text/plain, Size: 5265 bytes --]

Steve - I'm going to do a small rebuttal here.   You ask an interesting
question as we look back on history, but to be honest I. think the real
thing is that neither the Xerox folks nor the BTL folks in those days were
actually paying that much attention to each other AND they were considering
different problems.   It turns out, if they had "blended" their idea, maybe
the workstation world that would be birthed in the mid-1980s might have
been different.

On Sat, Jul 8, 2023 at 3:45 AM steve jenkin <sjenkin@canb.auug.org.au>
wrote:

> What struck me reading this is the estimated price (~$10K) to build an
> Alto, elsewhere I’ve seen $12K and 80 built in the first run.
>  [ a note elsewhere says $4,000 on 128KB of RAM. 4k-bit or 16-kbit chips?
> unsure ]
>
I'm pretty sure the first Alto's used Intel 1103A
<https://en.wikipedia.org/wiki/Intel_1103> (1Kx1), although the Intel 2106
(4Kx1) was coming on the scene - you would have to ask Roger Bates if he
remembers, as I believe that he designed the original memory boards for
Alto. That said, by the time of the second generation Alto's, Roger and
Chuck switched to a Mostek MK4116 - at least, that is my memory from the
ones we had a CMU a few years later.


> I’d suggest three reasons:
>
>         - The Consent Decree. AT&T couldn’t get into the Computer Market,
> only able to build computers for internal use.
>                 They didn’t need GUI PC’s to run telephone exchanges.
>
Put differently -- solving different markets.  Xerox was in the office
automation business - which was based on selling paper for their coping
machines.   Bob Taylor's real vision was that for all of the "paperless
office" comments of the day, he realized Xerox could sell way more paper if
it were easier to produce.

The Alto was thinking that at some time a currently very expensive piece of
equipment (the computer), would be cost-effective.  How would it be useful
for an office?

>
>         - Bell Labs management:
>                 they’d been burned by MULTICS and, rightly, refused the
> CSRC a PDP-10 in 1969.
>
>         - Nobody ’needed’ to save money building another DIY
> low-performance device.
>                 A home-grown supercomputer maybe :)
>
BTL was trying to figure out how to more efficiently write/program and
deploy them to aid in running a complex switching system - their business.
 Again the idea of a computing "utility" was often discussed.  Computers
cost big money to buy, deploy and operate. How do you use them better?



>
>
> It’s an accident of history that PARC could’ve, but didn’t, port Unix to
> the Alto in 1974.
>
Ouch ... not so fast.   First, the HW lacked an MMU.  I had to teach Roger
how they worked 5 years later when we built Magnolia and why they were a
good idea [I did the basic design of the Magonia MMU so we could run
UNIX].  I remember Roger telling me at the time, that Thacker and Lampson
didn't think they needed them if the computer was only being used for one
purpose. BTW: that was the same argument Jobs used a few years later when
Motorola offered them at a bargain price the optional MMU chip for the MAC
design they were developing because 68000 designers (Nick, Les, and team)
had used PDP-11s at Schumblege before then went to Motorola.  They knew
that not having an MMU was going to be a real problem, plus they had used a
UNIX box to help design their chip.

Second, Ken does not do his sabbatical to UCB until 1975.  While Butler
still had a fondness for UCB, there was not much interaction by then. At
the time, the primary computer at UCB was a CDC6600, and there was more
influence from "up the hill" on EECS from LBL than from PARC [and remember
that national labs like LBL are primarily using supers from CDC and later
Cray].  Also please remember that the "CS" types on the ARPANET are PDP-10
folks.  UCB does not have one.  In fact, the original ARPAnet connection
eventually to Ing70 was a VDH interface that ran down the hill from the IMP
at LBL, and that would not come for a few years yet.

PARC had much more influence at Stanford than at UCB. Stanford was on the
ARPANET [UCB was not for a long time yet] and Stanford was using PDP-10s.
PARC's Frankenstein MAXC was a PDP-10 clone with an Alto for its front end
instead of a PDP-11.




> By V7 in 1978, my guess it was too late because both sides had locked in
> ‘commercial’ positions and for PARC to rewrite code wasn’t justified: “If
> it ain’t Broke”…

Hmmmm, the PARC folks were rewriting code at this time.  The original Alto
is mostly BCPL and "Nova-Code" ( the original DG Nova microcode Bulter and
Chuck created).  By 1978, both Smalltalk and Cedar had been built at PARC.
Also, think about why Ethernet was created?   Metcalfe and Boggs proposed
it as a way to connect the different processors in a high-end Xerox copier.
It's used as a "network" that we know it was after they built it.   PUP and
things like the Press format [later XNS and Interpress come later].

So, I don't think it was anything more than just focus.  PARC was solving
one problem with the Alto and the BTL folks created UNIX to solve another.
ᐧ

[-- Attachment #2: Type: text/html, Size: 9162 bytes --]